Tongji Hospital Licenses Proprietary AI-Powered Dynamic Prediction System for High-Risk Sepsis Patients to Zaiyuan Bio for RMB 200,000

Recently, Tongji Hospital Affiliated to Tongji Medical College of Huazhong University of Science and Technology released a public notice on the transformation of scientific and technological achievements, proposing to transfer its independently developed“Predictive Support System for High-Risk Stratification in Sepsis Patients”Patent exclusive licensing conversion, with a proposed transfer amount of200,000 yuan, the Licensee is Wuhan Zaiyuan Biotechnology Co., Ltd.

Image from the official website of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology

The core technology of this patent revolves aroundDynamic Analysis of Physiological Data and Intelligent Risk Assessment in Sepsis PatientsBy leveraging innovative algorithms such as quantification of anomaly significance across different time periods, stratification of disease severity, and adaptive weighting of high-risk values, this approach addresses the limitation of traditional models in distinguishing the varying abnormal manifestations of sepsis at different stages. It significantly enhances the accuracy of high-risk prediction, providing intelligent technical support for early clinical warning and precise intervention.

SepsisIt is a systemic inflammatory response syndrome (SIRS) caused by an uncontrolled host immune response triggered by infection, formerly known as “septicemia” or “sepsis.” This is a critical condition with extremely rapid progression, characterized primarily byHigh Heterogeneity——There are significant differences among patients in their pathophysiological processes, the sequence of abnormal sign presentation, and the patterns of physiological parameter fluctuations.。

In clinical practice, patients often experience a rapid deterioration from a relatively stable condition within a short period, quickly progressing to high-risk critical conditions such as septic shock and multiple organ failure, which can even lead to death. Therefore,Early Identification of High-Risk Levels in Sepsis Patients and Timely Intervention, which is key to halting disease progression and saving patients' lives.

Currently, at both clinical and research levels, the prediction of high risk for sepsis generally relies on machine learning models, among whichRandom Forest ModelIt is a widely adopted technical solution. This method typically involves collecting physiological data from patients, such as body temperature, heart rate, blood pressure, and respiratory rate, to construct models for assessing patient risk levels.

However, constrained by the inherent complexity of sepsis, existing technical solutions face severe challenges in practical applications.Major Defects:Due to the high heterogeneity of sepsis, there are significant differences in the sequence and fluctuation patterns of abnormal signs among different patients, resulting in complex and variable manifestations of physiological data abnormalities at different time points.

More critically, the numerical characteristics of certain physiological indicators are quite similar during phases of patient deterioration (onset) and improvement. This similarity means that the distribution of physiological data may also be closely aligned during periods where abnormal manifestations differ. Such ambiguity in data features directly hinders the ability of existing random forest models to effectively distinguish between “high-risk” and “non-high-risk” states in sepsis.

Meanwhile, existing technologies fail to effectively address the insufficient distinguishability of time-segment data caused by variations in abnormal manifestations, resulting in low accuracy for predicting high-risk sepsis. This fails to meet the clinical practical needs for precise, dynamic, and early warning, thus clinicalThere is an urgent need for a predictive support system capable of dynamically amplifying differences in abnormal manifestations and enhancing data discriminability., to assist physicians in making more accurate clinical assessments and securing valuable time for patient treatment.

Addressing the critical challenges in clinical early warning for sepsis, this patented high-risk stratification prediction support system for sepsis patients proposes a novel dynamic analysis and intelligent prediction framework, delivering significant technical advantages and innovations:

First:Dynamic Quantification of the Importance of Physiological Indicators, automatically calculates the importance of each type of physiological indicator based on the severity of anomalies, time intervals, and the proportion of anomalous periods during abnormal episodes, accurately identifying key features that best reflect disease deterioration, thereby addressing the issue of unreasonable weight allocation in traditional methods.

Second:Multidimensional Assessment of Disease Severity, using the highest-importance indicators as feature categories and integrating logic such as differences in indicator importance, the number of abnormal categories, and correlation-based screening to objectively quantify disease severity during each critical period, thereby better aligning with the clinical characteristics of sepsis involving synchronous multi-organ and multi-indicator abnormalities.

Third:Adaptive Weighting of High-Risk Values, calculates high-risk values based on severity trend changes and time intervals, thereby dynamically amplifying physiological data features during periods of significant variation. This significantly enhances the model’s ability to distinguish between high-risk and non-high-risk states, overcoming the bottleneck of insufficient prediction accuracy in traditional random forest models.

Fourth:Full-Process Intelligent PredictionThe system integrates four core modules: data acquisition, importance analysis, severity calculation, and risk prediction. It can directly output standardized high-risk levels, providing real-time, objective, and interpretable decision support for clinical practice, making it particularly suitable for rapid early warning in emergency and critical care settings.

In summary, this systemAlgorithm Optimization, Multi-Scenario Adaptation, and Seamless Integration with Hospital Systems, with outstanding clinical value and high potential for widespread adoption, it boasts a broad market space in the intensive care units of medical institutions at all levels.

Global Sepsis Risk Prediction Has Been EstablishedTraditional Scoring Systems Coexist with Commercial AI Early Warning Systemsmarket landscape. In clinical practice, classic scoring systems such as qSOFA, SOFA, and SIRS remain the foundational screening tools; while widely adopted, they are limited by insufficient sensitivity, delayed results, and difficulty in dynamically tracking changes in patient condition.

Overseas markets are represented by mature commercial models such as Epic Systems’ SPM and Sepsis ImmunoScore,Leveraging deep integration with the electronic medical record (EMR) ecosystem to embed seamlessly into hospital workflows, it has achieved compliant implementation and large-scale adoption, primarily serving healthcare systems with robust IT infrastructure, such as those in North America.

In the domestic market, AI early-warning models are mostly in the stage of in-hospital development and small-scale validation. Nationally popular dynamic time-series products are relatively scarce. The overall landscape is characterized by rigid demand, insufficient supply, rapid technological iteration, and significant homogenized competition. Meanwhile, influenced by factors such as the level of healthcare informatization, data standardization, and adaptability to clinical workflows, there remains a substantial gap in high-quality, implementable products.

Epic Systems SPM (Sepsis Prediction Model) is a sepsis risk prediction clinical decision support tool embedded within Epic’s electronic health record system., a commercialized, embedded AI early-warning model that provides early sepsis identification and risk scoring for inpatients. It generates sepsis prediction scores by analyzing multidimensional data, including vital signs, laboratory tests, demographics, and comorbidities, and triggers clinical alerts to assist healthcare professionals in the early identification of suspected cases.

Deployed within the Epic EHR ecosystem, this model is primarily utilized by healthcare institutions in the United States, covering a large number of hospitals and health systems that use the Epic electronic health record system. It is one of the most widely adopted embedded commercial models for in-hospital sepsis early warning scenarios in North America. Its core features include deep integration with hospital information systems, ease of deployment, and direct alignment with clinical workflows. Institutions can adjust alert thresholds and notification rules according to their specific needs, ensuring compatibility with standard in-hospital diagnostic and treatment pathways.

In clinical applications, this modelStatic Weights, Batch ComputationAs the primary operational mode, it emphasizes risk identification based on full-cycle data. There is room for optimization in aspects such as the timing of alerts and differentiation of dynamic temporal patterns. Overall, it represents a mainstream embedded commercial competitor in the current field of in-hospital sepsis early warning systems.

Sepsis ImmunoScore is an AI-driven sepsis risk assessment and early warning tool developed by Prenosis., is a commercialized auxiliary diagnostic software that has obtained U.S. FDA De Novo marketing authorization. It primarily provides sepsis risk stratification for the next 24 hours in adult patients, helping clinicians identify potential high-risk cases earlier. The product performs comprehensive analysis based on multidimensional clinical and laboratory parameters, outputs risk assessment results, and serves as a decision-support reference for healthcare professionals.

The product adopts a deployment model that integrates with hospital information systems, enabling connectivity to electronic medical records (EMR) and related data platforms in healthcare institutions. Furthermore, through collaboration with the Roche Diagnostics Navify platform, it enhances accessibility and ease of use in clinical settings, seamlessly integrating into routine diagnostic and treatment workflows such as emergency care and inpatient wards, while aligning with standardized in-hospital screening and monitoring workflows.

In terms of market application, the product currently focuses on the U.S. healthcare market, serving hospitals and health systems with standardized data collection and informatization infrastructure, and targeting the critical clinical need for early sepsis identification. As an AI-based sepsis warning product that has completed regulatory compliance and achieved commercial deployment, it represents a relatively mature commercial solution in the international market, primarily providing technical support to medical institutions seeking to enhance the standardized management of sepsis.

This patented high-risk stratification prediction support system for sepsis patients precisely addresses current market pain points, withDynamic Anomaly Quantification, Temporal Severity Assessment, and Adaptive Weighting of High-Risk FactorsAs a core innovation, it differs from traditional static scoring systems and overseas fixed-weight models by better aligning with the clinical characteristics of sepsis—namely, its rapid progression and high heterogeneity—thereby significantly enhancing the accuracy and timeliness of risk identification.

Against the backdrop of policy-driven smart healthcare initiatives and tertiary hospitals’ enhanced focus on early warning systems for acute and critical conditions, this system can rapidly integrate with existing hospital monitoring devices and information systems. It is adaptable to multi-scenario applications including emergency departments, intensive care units (ICUs), and general wards, demonstrating strong feasibility for implementation and clinical utility.

Leveraging its differentiated technological advantages and addressing rigid clinical needs, this patent not only fills the market gap for dynamic intelligent early-warning tools in China but also enables the rapid development of commercial products through technology transfer. Consequently, it holds significant potential for deployment and promotion in critical care departments across medical institutions at all levels.